Efficient and noise robust photon-counting imaging with first signal photon unit method

Abstract

Efficient photon-counting imaging in low signal photon level is challenging, especially when noise is intensive. In this paper, we report a first signal photon unit (FSPU) method to rapidly reconstruct depth image from sparse signal photon counts with strong noise robustness. The method consists of acquisition strategy and reconstruction strategy. Different statistic properties of signal and noise are exploited to quickly distinguish signal unit during acquisition. Three steps, including maximum likelihood estimation (MLE), anomaly censorship and total variation (TV) regularization, are implemented to recover high quality images. Simulations demonstrate that the method performs much better than traditional photon-counting methods such as peak and cross-correlation methods, and it also has better performance than the state-of-the-art unmixing method. In addition, it could reconstruct much clearer images than the first photon imaging (FPI) method when noise is severe. An experiment with our photon-counting LIDAR system was conducted, which indicates that our method has advantages in sparse photon-counting imaging application, especially when signal to noise ratio (SNR) is low. Without the knowledge of noise distribution, our method reconstructed the clearest depth image which has the least mean square error (MSE) as 0.011, even when SNR is as low as −10.85 dB.